README.md

Segmentation of Blood Vessels in Retinal Images

Apply the U-Net to the segmentation of blood vessels in retinal images.

Env

• python 3.7
• pytorch 1.10.1
• torchmetrics 0.9.0

Dataset

DRIVE: Digital Retinal Images for Vessel

Network

Original paper by Olaf Ronneberger, Philipp Fischer, Thomas Brox:

U-Net: Convolutional Networks for Biomedical Image Segmentation

Usage

Train the network:

python train.py

Predict in the test dataset:

python predict.py

you would expect the get the accuracy, precision, recall, specificity and ROC curve in the terminal.

Here is the ROC curve.

and here is the result (not very satisfactory ...):